In my language the ability to abstract is part of the definition of intelligence and that’s something no current “AI” can do. To be considered intelligent the program would have to be able to derive solutions for previously unknown problems from it’s current knowledge only.